SU‐E‐T‐497: A Method for Automatic Commissioning of a GPU‐Based Monte Carlo Code for Clinical Photon Beam Dose Calculation

Z. Tian, R. Townson, Y. Graves, X. Jia, S. Jiang

Research output: Contribution to journalArticlepeer-review

Abstract

Purpose: Monte Carlo (MC) method is recognized as the most accurate method for dose calculation. The commissioning of beam models is crucial for the clinical implementation of a MC code. We propose an automatic commissioning method for our GPU‐based MC dose engine (gDPM) using a source model based on the concept of phase‐space‐let (PSL). Methods: A PSL contains a group of particles that are of the same type and close in space and energy. PSLs for a reference phase‐space file are first generated, and dose for each PSL is pre‐calculated in water. Weighting factor of each PSL is adjusted to fine tune the fluence distribution and energy spectrum, so that the corresponding calculated dose matches the standard measured dose in water such as PDD and dose profiles. This is essentially an underdetermined least‐square minimization problem. Therefore, we add symmetric and smooth regularizations to the optimization problem, assuming that the dose profiles are symmetric and smooth. A split Bregman method is adopted to solve this optimization problem. Results: The phase‐space file of a Varian TrueBeam 6MV beam was used to generate PSLs for all 6MV beams. First, for a simulation study, the ‘measurement’ data was obtained from MC dose calculation using a set of field‐size‐dependent phase‐space files of a Siemens 6MV beam. The PDD and dose profiles calculated from the commissioned beam model agree well (within 1%) with the ‘measurement’ data. Second, realistic clinical data of Varian, Siemens, and Elekta machines were also used to test our method and achieved similar commissioning accuracy of less than 1% dose difference. Conclusion: Using the PSL source model, the proposed commissioning method can automatically fine tune the fluence distribution and energy spectrum of the reference phase‐space file to match measured data for clinical beam to be commissioned, relieving clinical MC users from conventional cumbersome commissioning process.

Original languageEnglish (US)
Pages (from-to)319
Number of pages1
JournalMedical physics
Volume40
Issue number6
DOIs
StatePublished - Jun 2013

ASJC Scopus subject areas

  • Biophysics
  • Radiology Nuclear Medicine and imaging

Fingerprint

Dive into the research topics of 'SU‐E‐T‐497: A Method for Automatic Commissioning of a GPU‐Based Monte Carlo Code for Clinical Photon Beam Dose Calculation'. Together they form a unique fingerprint.

Cite this